Restore skipping color correction for byte rasters

[James Santucci]

Problem description

Imagery already in the byte range doesn't need correction, and also, our histogram stretch seems to wash it out / introduce artifacts.

We used to do in ColorCorrect.scala

// if data is a byte raster (0, 255) don't do anything
// else if stats are available, clip assuming a normal distribution
// else use the histogram's min/max
if (imin >= 0 && imax <= 255) {
  iMaxMin(index) = (0, 255)
} else {
  (imin, imax, statsOption) match {
    case (_, _, Some(stats)) =>
      // assuming a normal distribution, clips 2nd and 98th percentiles of values
      val newMin =
        if (disableCorrection) imin
        else stats.mean + (stats.stddev * -2.05)
      val newMax =
        if (disableCorrection) imax
        else stats.mean + (stats.stddev * 2.05)
      // assume non-negative values, otherwise visualization is weird
      // I think this happens because the distribution is non-normal
      iMaxMin(index) =
        (if (newMin < 0) 0 else newMin.toInt, newMax.toInt)
    case (min, max, _) => iMaxMin(index) = (min, max)
  }
}

But when we simplified color correction, we lost the logic to skip correction for byte rasters.

Replication Steps

• upload a tif in the 0 - 255 range
• look at it
• look at it with disableAutoCorrection=true
• feel sad

Expected Behavior or Output

all the imagery looks great all the time

[James Santucci]

Do azavea/raster-foundry-platform#1219 at the same time -- they're both normalization errors that will be easy to track down